The UUV Obstacle Avoidance and Navigation Method Based on Fuzzy Control-Enhanced SAC

Obstacle avoidance and navigation is a crucial and challenging task in the exploration of unmanned underwater vehicle (UUV) in unknown underwater environments. In recent years, researchers have explored the combination of deep reinforcement learning (DRL) to address obstacle avoidance and navigation for UUV However, DRL struggles to utilize the limited samples collected in real-world environment. That makes it challenging to apply DRL methods into real underwater environments. This paper introduces the FZSAC algorithm, utilizing fuzzy control to integrate a small number of rules, ensuring that UUV can collect better samples in real- time environments. This enhances the online updating efficiency of the SAC algorithm, addressing the issue of slow convergence. In the pybullet simulation environment, we experimentally validate that FZSAC accelerates model convergence and accomplishes obstacle avoidance and navigation tasks in both static and dynamic environments. Furthermore, the method exhibits favorable performance across multiple test environments and shows good generalization.